Unverified50% confidenceFactExact time
Over the past two years, the score gaps among leading models on mainstream benchmarks have narrowed from substantial differences to single-digit percentage points
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/6/2026
First Seen
Valid until: 1/4/2027
Sources
Related Entities
Related Claims
UnverifiedThe gap between flagship models from multiple laboratories on mainstream evaluations has narrowed to single-digit percentage points.70% similarUnverified斯坦福大学HAI研究所2024年度AI指数报告显示，顶级闭源模型在MMLU基准上的最高分与第五名之间的差距，已从2022年的约8个百分点收窄至不足2个百分点67% similarVerifiedOn mainstream evaluations such as MMLU, HumanEval, and GSM8K, the score gaps among leading large models are often only a few percentage points.63% similarUnverified当综合评估指数中顶尖模型得分接近天花板时会出现基准饱和问题，60分对59分的差距在统计学上通常处于误差范围之内62% similarUnverifiedbenchmark分数与实际开发体验之间存在显著落差，模型可能在标准测试中领先却在真实项目中表现平平61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/980311API
curl https://kongchang.com/api/v1/knowledge/claims/980311MCP
get_claim(id=980311)